open-nomad/client/allocrunner/taskrunner/stats_hook.go

// Copyright (c) HashiCorp, Inc.
// SPDX-License-Identifier: MPL-2.0

package taskrunner

import (
	"context"
	"sync"
	"time"

	hclog "github.com/hashicorp/go-hclog"
	"github.com/hashicorp/nomad/client/allocrunner/interfaces"
	cstructs "github.com/hashicorp/nomad/client/structs"
	"github.com/hashicorp/nomad/nomad/structs"
	bstructs "github.com/hashicorp/nomad/plugins/base/structs"
)

// StatsUpdater is the interface required by the StatsHook to update stats.
// Satisfied by TaskRunner.
type StatsUpdater interface {
	UpdateStats(*cstructs.TaskResourceUsage)
}

// statsHook manages the task stats collection goroutine.
type statsHook struct {
	updater  StatsUpdater
	interval time.Duration

	// cancel is called by Exited
	cancel context.CancelFunc

	mu sync.Mutex

	logger hclog.Logger
}

func newStatsHook(su StatsUpdater, interval time.Duration, logger hclog.Logger) *statsHook {
	h := &statsHook{
		updater:  su,
		interval: interval,
	}
	h.logger = logger.Named(h.Name())
	return h
}

func (*statsHook) Name() string {
	return "stats_hook"
}

func (h *statsHook) Poststart(_ context.Context, req *interfaces.TaskPoststartRequest, _ *interfaces.TaskPoststartResponse) error {
	h.mu.Lock()
	defer h.mu.Unlock()

	// This shouldn't happen, but better safe than risk leaking a goroutine
	if h.cancel != nil {
		h.logger.Debug("poststart called twice without exiting between")
		h.cancel()
	}

	// Using a new context here because the existing context is for the scope of
	// the Poststart request. If that context was used, stats collection would
	// stop when the task was killed. It makes for more readable code and better
	// follows the taskrunner hook model to create a new context that can be
	// canceled on the Exited hook.
	ctx, cancel := context.WithCancel(context.Background())
	h.cancel = cancel
	go h.collectResourceUsageStats(ctx, req.DriverStats)

	return nil
}

func (h *statsHook) Exited(context.Context, *interfaces.TaskExitedRequest, *interfaces.TaskExitedResponse) error {
	h.mu.Lock()
	defer h.mu.Unlock()

	if h.cancel == nil {
		// No stats running
		return nil
	}

	// Call cancel to stop stats collection
	h.cancel()

	// Clear cancel func so we don't double call for any reason
	h.cancel = nil

	return nil
}

// collectResourceUsageStats starts collecting resource usage stats of a Task.
// Collection ends when the passed channel is closed
func (h *statsHook) collectResourceUsageStats(ctx context.Context, handle interfaces.DriverStats) {

MAIN:
	ch, err := h.callStatsWithRetry(ctx, handle)
	if err != nil {
		return
	}

	for {
		select {
		case ru, ok := <-ch:
			// if channel closes, re-establish a new one
			if !ok {
				// backoff if driver closes channel, potentially
				// because task shutdown or because driver
				// doesn't implement channel interval checking
				select {
				case <-time.After(h.interval):
					goto MAIN
				case <-ctx.Done():
					return
				}
			}

			// Update stats on TaskRunner and emit them
			h.updater.UpdateStats(ru)

		case <-ctx.Done():
			return
		}
	}
}

// callStatsWithRetry invokes handle driver Stats() functions and retries until channel is established
// successfully.  Returns an error if it encounters a permanent error.
//
// It logs the errors with appropriate log levels; don't log returned error
func (h *statsHook) callStatsWithRetry(ctx context.Context, handle interfaces.DriverStats) (<-chan *cstructs.TaskResourceUsage, error) {
	var retry int

MAIN:
	if ctx.Err() != nil {
		return nil, ctx.Err()
	}

	ch, err := handle.Stats(ctx, h.interval)
	if err == nil {
		return ch, nil
	}

	// Check if the driver doesn't implement stats
	if err.Error() == cstructs.DriverStatsNotImplemented.Error() {
		h.logger.Debug("driver does not support stats")
		return nil, err
	}

	// check if the error is terminal otherwise it's likely a
	// transport error and we should retry
	if re, ok := err.(*structs.RecoverableError); ok && re.IsUnrecoverable() {
		h.logger.Debug("failed to start stats collection for task with unrecoverable error", "error", err)
		return nil, err
	}

	// We do not warn when the plugin is shutdown since this is
	// likely because the driver plugin has unexpectedly exited,
	// in which case sleeping and trying again or returning based
	// on the stop channel is the correct behavior
	if err == bstructs.ErrPluginShutdown {
		h.logger.Debug("failed to fetching stats of task", "error", err)
	} else {
		h.logger.Error("failed to start stats collection for task", "error", err)
	}

	limit := time.Second * 5
	backoff := 1 << (2 * uint64(retry)) * time.Second
	if backoff > limit || retry > 5 {
		backoff = limit
	}

	// Increment retry counter
	retry++

	time.Sleep(backoff)
	goto MAIN
}

func (h *statsHook) Shutdown() {
	h.mu.Lock()
	defer h.mu.Unlock()

	if h.cancel == nil {
		return
	}

	h.cancel()
}